Networked computer systems may be used to offload workloads to multiple devices, store data on multiple devices, and so forth. Thus, it can be important to effectively manage how the system reacts when a device within the system fails. Failure management algorithms can establish and monitor certain failure criteria to determine when a device is failing. Failure management algorithms may use fixed failure criteria on a per-system basis that cannot be dynamically adjusted based on failure data from other devices or systems in the networked computer system.